Skip to Content

Diagnosing Disease with a Snapshot

Many genetic conditions come with clues in a person’s face, and new technology can help doctors diagnose them.
December 6, 2016

Not long ago, a geneticist in Delaware was doing what geneticists do, trying to arrive at a diagnosis for a patient with undiagnosed disease.

The physician, Karen Gripp, chief of the Division of Medical Genetics at A.I. duPont Hospital for Children in Wilmington, Delaware, noted that her patient, a toddler girl, had a high forehead, thick eyebrows, and a long philtrum—the space between the nose and the lip. She experienced seizures and had coarse, curly hair, though that hardly seemed noteworthy due to her African-American ethnicity. A cleft palate plus an arachnoid cyst on her spinal cord had already been repaired by the time Gripp was called to see her.

The symptoms didn’t line up neatly with any conditions Gripp had considered.

Many genetic conditions have a “face”—a distinctive constellation of features that provides a clue to a potential diagnosis. But identifying disease by observing a patient’s features, a practice known as dysmorphology, is one of geneticists’ biggest challenges. The most skilled dysmorphologists tend to be older doctors who’ve been around the block, which makes sense; the more patients you see over a lifetime, the more features you observe. But even the most experienced practitioners haven’t laid eyes on every disease out there. 

As part of her detective work, Gripp asked permission to take a picture of the patient and upload it to Face2Gene, a facial recognition software tool that can aid in rare disease diagnosis. Face2Gene compares pictures of a patient’s face with those of disease composites and returns a series of potential diagnoses, from most plausible to least.

For Gripp, the technology proved illuminating. Hajdu-Cheney syndrome was one of Face2Gene’s top suggestions for the toddler. But an exome analysis of the child’s protein-coding genes revealed a mutation implicated in lateral meningocele syndrome, a condition with which Gripp was more than familiar. She had written the academic paper and the National Library of Medicine’s GeneReview about it. 

Cleft lips are not a feature of lateral meningocele syndrome, so Gripp had not initially considered it as a diagnosis. But its similarity to Hajdu-Cheney syndrome helped point her to it. So few patients have this diagnosis that the child’s future is not clear. She will survive, although she’ll be closely monitored for possible neurologic problems as well as scoliosis, which has been closely associated with the condition.

“I kept looking at conditions with cleft lip, but that was a total red herring,” says Gripp, who believes that the cleft is unrelated to the girl’s genetic disorder. “You only recognize what you know. Face2Gene gets me to go down the list and say, ‘Hmm, did I consider everything?’”

FDNA, Face2Gene’s parent company, began six years ago after the Israeli cofounders sold their previous facial recognition company,, to Facebook. That technology is able to differentiate specific individuals by being “trained” on multiple images of that person. Face2Gene’s technology, in contrast, identifies a pattern that is common to a group of people that have the same syndrome; establishing that common denominator allows the software to create a composite image associated with a condition.

When CEO Dekel Gelbman was hired in 2010, he met with multiple practitioners and quickly realized that facial recognition could help reduce the burden of undiagnosed disease. 

Of the more than 7,000 known genetic syndromes, Face2Gene estimates that up to half have a distinct facial pattern that can be learned and used for diagnosis. Down syndrome, for example, is among the most common and is therefore easier to diagnose. But rarer conditions can prove more challenging. “Imagine you’re a geneticist and you have your own repository of images in your brain according to what you’ve seen and been trained on,” says Gelbman. “You can’t know everything. Your ability to conjure up an image or try to compare these patterns in your head is limited. So how do you democratize this?”

Face2Gene crowdsources data from geneticists.  Though the technology is free for providers, Gelbman envisions charging pharmaceutical companies for access to aid in drug discovery and development. The more people who input data into Face2Gene, the more the system learns which facial features are associated with which syndromes. Gelbman says that 60 percent of clinical geneticists and genetic counselors worldwide have used the technology.

The mobile app automatically photographs a patient, uploads that photo to a server, and analyzes the facial features within seconds to generate a list of syndromes that match identified similarities. Each syndrome is accompanied by information from the London Medical Databases, which curates syndromes and maintains a collection of images of dysmorphology.

Users can superimpose their patient’s face onto a disease prototype’s face and click on a heat map that reveals those facial areas that most closely align with a representative image of a particular syndrome. They can also tag features that are present in their patient, which recalculates the list of potential syndromes. 

Up next for FDNA is continued work on trying to determine facial features for various medical conditions, notably autism. “We want to know if we can find a face for fragile X,” says Gelbman, referencing the most common inherited cause of intellectual disability in boys. “We’ll see what we can find.”

Keep Reading

Most Popular

Large language models can do jaw-dropping things. But nobody knows exactly why.

And that's a problem. Figuring it out is one of the biggest scientific puzzles of our time and a crucial step towards controlling more powerful future models.

The problem with plug-in hybrids? Their drivers.

Plug-in hybrids are often sold as a transition to EVs, but new data from Europe shows we’re still underestimating the emissions they produce.

How scientists traced a mysterious covid case back to six toilets

When wastewater surveillance turns into a hunt for a single infected individual, the ethics get tricky.

Google DeepMind’s new generative model makes Super Mario–like games from scratch

Genie learns how to control games by watching hours and hours of video. It could help train next-gen robots too.

Stay connected

Illustration by Rose Wong

Get the latest updates from
MIT Technology Review

Discover special offers, top stories, upcoming events, and more.

Thank you for submitting your email!

Explore more newsletters

It looks like something went wrong.

We’re having trouble saving your preferences. Try refreshing this page and updating them one more time. If you continue to get this message, reach out to us at with a list of newsletters you’d like to receive.